Poly-Glot AI Workspace

MCP serverAI & models

Your AI can start from a ready-made prompt in any of 38 languages instead of writing one from scratch. poly-glot-ai-workspace is an integration that adds a library of more than 1,000 prompt templates plus seven tools for working with them.

Available today. Use it from your connected AI after setup.

After adding it, tell your AI what you are working on and which language you need, then ask it to find a matching template. Try a few templates to see which wording works best for your task.

Then ask your AI: use Poly-Glot AI Workspace to get language options

What your AI can do with it

  • Draw on over 1,000 ready-made prompt templates
  • Work with prompts in 38 languages
  • Find a starting prompt for a task in the language you need
  • Use seven built-in tools while working with prompts

From the project's README

As published by hmoses/poly-glot-ai-workspace in README.md.

Multilingual MCP tool platform: 1,000+ prompt templates · 35 languages · Compare Mode · BYOM · Embedded GUI

Remote MCP server with an embedded MCP Apps / ChatGPT Apps SDK GUI, server-side entitlements, Apple-backed subscriptions, and Bring Your Own Model (BYOM) support.

MCP Endpoint

https://br-steep-leaf-ae2o29qz-mcp.compute.c-2.us-east-2.aws.neon.tech/mcp

Transport: Streamable HTTP

Tools

search_templates

Find Poly-Glot prompt templates by keyword, goal, or plan. Results include whether each template is currently locked for the account.

Parameters: query (string), goal (string), plan ("free" | "pro"), limit (1–24, default 12), uiLanguage (string, default "EN")

get_template

Get a template's fields and prompt body. The source prompt body is returned only when the account is entitled to use the template.

Parameters: name (string, required), uiLanguage (string, default "EN")

build_prompt

Fill an entitled Poly-Glot template with user values. First eligible use starts the 3-day trial. During the active trial, all 1,000+ templates are unlocked; after the trial, template bodies lock unless the account is Pro.

Parameters: name (string, required), values (object), uiLanguage, inputLanguage, outputLanguage (strings, default "EN")

prepare_compare

Prepare one canonical prompt for 2+ AI providers so the user can compare answers side-by-side. Does not call third-party models on the user's behalf.

Parameters: name (string), prompt (string), values (object), providers (array of "chatgpt" | "claude" | "gemini" | "perplexity" | "grok" | "copilot" | "mistral", min 2), uiLanguage, inputLanguage, outputLanguage

get_language_options

Return the 35 supported Poly-Glot UI, input, and AI output languages. Language selection never changes entitlement.

Parameters: uiLanguage (string, default "EN")

get_subscription_status

Return the current Poly-Glot trial or Pro entitlement state and current pricing.

Parameters: none

open_workspace

Open the interactive Poly-Glot template browser and prompt editor with subscription-aware locked states. In UI-capable MCP hosts (ChatGPT), renders the full embedded GUI inside the conversation.

Parameters: query (string), uiLanguage (string, default "EN")

transcribe_audio

Transcribe audio content and return text. Supports language detection and forced source language. Requires entitlement.

Parameters: audioUrl (string, required), sourceLanguage (string), uiLanguage (string, default "EN")

detect_language

Detect the language of a text snippet. Returns ISO code and confidence score.

Parameters: text (string, required), uiLanguage (string, default "EN")

translate_text

Translate text between any of the 35 supported Poly-Glot languages.

Parameters: text (string, required), from (string), to (string, required), uiLanguage (string, default "EN")

localize_text

Localize text for a target locale — adapts tone, units, date formats, and cultural references beyond simple translation.

Parameters: text (string, required), targetLocale (string, required), context (string), uiLanguage (string, default "EN")

get_custom_model_capabilities

Return supported BYOM adapter modes, credential policy, network restrictions, and notes about localhost access.

Parameters: none

validate_custom_model

Validate a developer-supplied model endpoint. Checks HTTPS, SSRF, and optionally probes the model. API keys are transient and never persisted or echoed.

Parameters: adapterMode ("openai-compatible" | "custom-rest"), baseUrl (string), endpoint (string), model (string), apiKey (string, transient), probe (boolean)

run_custom_model

Run a Poly-Glot prompt against a developer-supplied model endpoint. Applies language instructions and respects entitlement checks.

Parameters: adapterMode ("openai-compatible" | "custom-rest"), baseUrl/endpoint (string), model (string), apiKey (string, transient), prompt (string, required), system (string), temperature, maxTokens, uiLanguage, inputLanguage, outputLanguage

prepare_custom_compare

Build a Compare Mode plan containing built-in providers and developer-supplied custom model descriptors. Credentials are supplied at execution time only.

Parameters: prompt (string, required), builtinProviders (array), customModels (array of { label, adapterMode, baseUrl/endpoint, model }), uiLanguage, inputLanguage, outputLanguage

Pricing and access

  • 3-day free trial — full access to all 1,000+ templates, Compare Mode, unlimited sends, BYOM, and language tools.
  • After trial — 1 free single-AI send per day; Compare Mode and premium features lock.
  • Pro Monthly — $9.99/month for unlimited access.
  • Pro Annual — $79.99/year (Save 33%) for unlimited access.

Features

  • 1,022 prompt templates — 25 designated Free, 997 designated Pro; all 1,022 are unlocked during the active trial
  • 35 languages — independent UI, input, and output language controls
  • Compare Mode — same prompt across ChatGPT, Claude, Gemini, Perplexity, Grok, Copilot, Mistral
  • Bring Your Own Model (BYOM) — connect OpenAI-compatible or custom REST HTTPS endpoints; credentials transient, never persisted
  • Embedded GUI — interactive workspace widget in ChatGPT and MCP Apps-capable hosts
  • Server-side entitlements — Apple-backed subscription verification via Neon Postgres

Entitlement tiers

FeatureFree / not startedTrial (3 days)Expired (post-trial)Pro
Browse & search templates
Language options & status
Template bodiesFree templates only✅ All❌ Locked✅ All
Send to 1 AI✅ starts trial✅ Unlimited✅ 1/day✅ Unlimited
🔀 Compare Mode✅ starts trial
Language processing
BYOM execution

Endpoints

SurfaceURL
MCP (Streamable HTTP)https://br-steep-leaf-ae2o29qz-mcp.compute.c-2.us-east-2.aws.neon.tech/mcp
Healthhttps://br-steep-leaf-ae2o29qz-mcp.compute.c-2.us-east-2.aws.neon.tech/
Entitlement APIhttps://br-steep-leaf-ae2o29qz-entitlements.compute.c-2.us-east-2.aws.neon.tech/

Connect

Add this URL as a remote MCP server in any MCP-compatible client:

https://br-steep-leaf-ae2o29qz-mcp.compute.c-2.us-east-2.aws.neon.tech/mcp

Works with: ChatGPT (with GUI) · Claude · Goose · Cursor · any Streamable HTTP MCP client

Local development

npm install
npm run sync
npm run check
npm start

Local MCP endpoint: http://localhost:8787/mcp

Links

Tools it offers (15)

What this server listed when ahel dialed its public endpoint in Sep 2026, with no key and no account of yours. The names are the server’s own.

  • get_language_options
  • get_subscription_status
  • open_workspace
  • search_templates
  • get_template
  • build_prompt
  • prepare_compare
  • get_custom_model_capabilities
  • validate_custom_model
  • run_custom_model
  • prepare_custom_compare
  • transcribe_audio
  • detect_language
  • translate_text
  • localize_text

Signals

Last commit
Sep 2026
Advanced
Delivery
poly-glot-ai-workspace MCP server → your ahel gateway (mcp.ahel.ai) → every connected AI client.
Catalog kind
mcp-server
Gateway key
io-github-hmoses-poly-glot-ai-workspace
Source
github.com/hmoses/poly-glot-ai-workspace
Hosted endpoint
https://br-steep-leaf-ae2o29qz-mcp.compute.c-2.us-east-2.aws.neon.tech/mcp